When I train the Harvard dataset, the loss is very low from the first epoch (loss=0.0042 when the dataset was just loaded for training, and it has dropped to 0.0008 by the end of the first epoch), and the PSNR will reach an outrageous 65dB. How can this situation be resolved? How did you train the Harvard dataset?
When I train the Harvard dataset, the loss is very low from the first epoch (loss=0.0042 when the dataset was just loaded for training, and it has dropped to 0.0008 by the end of the first epoch), and the PSNR will reach an outrageous 65dB. How can this situation be resolved? How did you train the Harvard dataset?